Building TinyML Pipelines Training in Cork, Ireland

The Building TinyML Pipelines Training is a hands-on, practical program focused on designing, training, deploying, and operating machine learning models on ultra-low-power, resource-constrained devices such as microcontrollers and edge sensors.

Participants will learn how to build end-to-end TinyML pipelines—from data collection and model training to optimization, deployment, and lifecycle management—enabling real-time intelligence at the edge for IoT, industrial, wearable, and embedded systems.

Introduction to TinyML & Edge AI

• What is TinyML and why it matters

• TinyML vs traditional edge and cloud ML

• Benefits of on-device inference

• Common TinyML use cases and limitations

TinyML Pipeline Architecture

• End-to-end TinyML workflow

• Data → model → deployment → inference loop

• Cloud + edge collaboration patterns

• Designing for low power and low memory

Sensor Data Collection & Labeling

• Types of sensors for TinyML

• Streaming and batch data capture

• Data labeling strategies

• Handling noise and sensor drift

Data Preprocessing for TinyML

• Feature extraction for time-series and signals

• Windowing and segmentation

• Normalization and scaling

• Reducing data size for embedded use

Model Selection for TinyML

• Choosing lightweight model architectures

• Trade-offs between accuracy and footprint

• Classical ML vs small neural networks

• Model complexity vs device constraints

Training TinyML Models

• Training workflows on PCs/cloud

• Validation and testing strategies

• Preventing overfitting on small datasets

• Performance metrics for TinyML

Model Optimization & Compression

• Quantization techniques

• Pruning and weight sharing

• Reducing memory and compute footprint

• Balancing latency, accuracy, and power

Deployment to Microcontrollers

• Converting models for embedded deployment

• Integrating models into firmware

• Memory mapping and resource allocation

• Build and flash workflows

On-Device Inference & Optimization

• Running inference on constrained hardware

• Measuring latency and power usage

• Real-time vs batch inference

• Debugging on-device ML behavior

Edge-to-Cloud Integration

• Sending inference results to the cloud

• Remote monitoring and logging

• Device fleet management concepts

• Feedback loops for model improvement

Updating & Managing TinyML Pipelines

• Model update strategies

• Versioning and rollback

• Continuous improvement cycles

• Managing model drift at the edge

Security & Reliability for TinyML

• Securing models and firmware

• Protecting intellectual property

• Safe and reliable inference

• Fault tolerance on embedded devices

Hands-on Exercises

Summary and Conclusion